Bitcoin rose 42.7% in the third quarter (Q3) of 2026 to $83,570 (Glassnode, 21shares, as of September 30, 2026). The question for an investment committee is who bought it and whether that buyer persists. Bitcoin ETF inflows supplied the answer: US spot exchange-traded funds (ETFs) overtook bitcoin treasury companies as the main source of demand.
Key takeaways
- US spot bitcoin exchange-traded funds (ETFs) added 79,800 bitcoin ($6.2 billion) in the third quarter of 2026, nearly twice the 41,300 bitcoin ($2.9 billion) mined, after $4.8 billion of outflows in the second quarter (Bloomberg, Glassnode, 21shares, as of September 30, 2026).
- Long-term holders reduced their balances by 248,000 bitcoin ($17.4 billion) in Q3 2026, nearly three times ETF purchases, while miner income exceeded estimated total mining costs by 6% in September 2026 after 10 months below breakeven (Glassnode, Cambridge Bitcoin Electricity Consumption Index, as of September 30, 2026).
- Transaction fees covered 0.67% of bitcoin miner income in Q3 2026, and at September 2026 prices, the 2028 halving, the scheduled cut in the block reward, would leave rewards covering 80% of power costs (21shares, Glassnode, Cambridge Bitcoin Electricity Consumption Index, as of September 30, 2026).
Bitcoin ETF inflows reached nearly twice newly mined supply
US spot bitcoin ETFs added 79,800 bitcoin ($6.2 billion) in Q3 2026, against 41,300 bitcoin ($2.9 billion) newly mined (Glassnode, 21shares, as of September 30, 2026). The same funds recorded $4.8 billion of outflows in the second quarter (Q2). That swing changed who was buying. However, ETF buying was heavily concentrated in two periods, August 17-27 and September 21-25, accounting for 91% of Q3 inflows measured in bitcoin.
Bitcoin treasury companies, listed companies that hold bitcoin on their balance sheets, moved the other way. Their buying fell from $7.1 billion in Q2 to roughly $1 billion in Q3, and Strategy added only 1,700 bitcoin ($240 million) after selling $349 million in July and August (Strategy filings, 21shares, Q3 2026). Together, ETFs and treasury companies added about $7.2 billion, 2.4 times the value of newly mined bitcoin, up from 0.8 times in Q2 (21shares, Glassnode, Q2 to Q3 2026).
ETF dollar flows are from Bloomberg; other dollar values are summed at the price on the day of each flow. Source: 21shares, Glassnode, Bloomberg, Strategy filings, CoinGecko. Data as of September 30, 2026.
Long-term holders sold nearly three times what ETFs bought
Long-term holders, defined as wallets whose coins have not moved in more than 155 days, reduced their balances by 248,000 bitcoin ($17.4 billion) in Q3 2026 (Glassnode, as of September 30, 2026). The same group had absorbed the downturn. After the October 10, 2025 crash, long-term holders added $31 billion between November and June 2026, 1.75 times the $17.6 billion added by treasury companies, while ETFs shed $7.6 billion (21shares, Glassnode, Bloomberg, November 2025 to June 2026).
Selling into the rally produced gains. Investors overall realized $9.3 billion in net profits in Q3, the first quarterly net gain of 2026, after $61.6 billion of net realized losses in the first half. Q2 and Q3 combined still show a net realized loss of $12.3 billion (Glassnode, as of September 30, 2026).
Price has moved well ahead of holder cost but remains below its historical norm. Holders' average acquisition cost, or realized price, rose only 0.9% to $53,637 in Q3. The market value to realized value (MVRV) ratio, which divides market price by that average cost, reached 1.56x against a median of 1.71x since 2013 (21shares, Glassnode, 2013-2026). Coins that last moved one to two years ago were acquired at about $97,000, and the price sits at 0.86 times that cost. 21shares analysis identifies that level as one where selling pressure could build, and roughly 30% of supply remains at a loss. Bitcoin also remains 26.7% below its price a year earlier (Glassnode, September 30, 2025 to September 30, 2026).
Futures trading dominated as spot volume fell faster
Futures accounted for 88.2% of Q3's $4.0 trillion in bitcoin trading volume, up from 86.7% in Q2. Spot volume fell 23% quarter-on-quarter and 53% year-on-year to $470 billion, while futures volume fell 7% and 29% to $3.49 trillion (Glassnode, as of September 30, 2026). Spot buying therefore made up a smaller share of activity than the price move suggests, meaning ETF flows carry more weight in reading demand.
More transactions did not mean more value moving onchain
Onchain activity, recorded directly on the Bitcoin blockchain, rose in count but fell in value. Transactions increased 46.8% year on year to 62.06 million in Q3 2026, while adjusted transfer value fell 48.7% to $1.16 trillion and was flat quarter on quarter (Glassnode, as of September 30, 2026).
Transfers were smaller and fewer addresses were active. The median transfer averaged $62, down from $152 a year earlier. Daily active addresses fell 11.5% year on year to 639,381, despite rising by 1.9% over the last quarter. The realized capitalization to transfer volume (RVT) ratio, which compares value held with value moved, fell 17.8% in the quarter to 91.3 but remained twice its level from a year ago, as the value transferred onchain fell while the value held stayed roughly flat. Finally, Public Lightning network capacity also dropped 19.3% in Q3 to 3,767 bitcoin ($310 million). For reference, this figure captures only visible channel capacity and does not measure payment volume or private channels (Glassnode, as of September 30, 2026).
YTD figures compare January-September 2026 with the same period in 2025, except RVT and Lightning capacity, which measure the change from December 31, 2025 to September 30, 2026.
Miners moved back above breakeven after 10 months
September 2026 miner income exceeded estimated total mining costs by 6%, ending 10 months below breakeven. Estimated costs are calculated from Cambridge network power at $0.05 per kilowatt-hour with a 1.5 multiplier, and income recovered to roughly its one-year average (Glassnode, Cambridge Bitcoin Electricity Consumption Index, as of September 30, 2026). Margins are positive but thin.
Miners absorbed a severe price decline before that. From October 2025 to June 2026, bitcoin's average price fell 45%, yet miner income slipped only from 1.6 times to 1.3 times the network's power bill, because weaker miners shut down and mining difficulty, the effort required to mine a block, fell. Mining power dropped 17% between October 2025 and July 2026 (21shares, Glassnode, Cambridge Bitcoin Electricity Consumption Index). However, some of that resilience comes from Cambridge’s model, which excludes mining machines that would lose money at the assumed Bitcoin price, naturally tilting estimates towards breakeven. Q3 miner revenue fell 40% year on year in dollars but only 3.6% in bitcoin terms, while average hashrate, the computing power securing the network, fell 4.6% to 918 exahashes per second (EH/s) (Glassnode, 21shares on Dune, as of September 30, 2026).
Users fund less than 1% of bitcoin's security budget
Transaction fees covered 0.67% of the network's $2.96 billion Q3 security cost, so newly issued bitcoin funds almost all of it. Fees rose 6.5% in the quarter to $19.9 million but fell 54.2% year on year, while holder dilution from issuance, the cost holders bear when miners are paid in new bitcoin, reached $2.94 billion. Issuance is adding to supply at an annualized rate of roughly 0.8% (Glassnode, 21shares on Dune, as of September 30, 2026).
Mining also became more concentrated. The top three mining pools, groups that combine computing power, produced 60.0% of blocks in Q3, up from 58.2% in Q2, though their January to September share remained below the same period in 2025. No single pool has majority control, and the largest holds 25.5% (Glassnode, 21shares on Dune, as of September 30, 2026).
The 2028 halving will test miner margins
Bitcoin miners have consistently earned more than their power bill, but the 2028 halving will put that resilience to the test. Since 2018, miner income has never fallen below 1.23 times the network’s power bill, and it stood at 1.59 times in September 2026, or 1.06 times full mining cost. At September’s price and network computing power, the halving would leave rewards covering just 80% of the power bill. Returning to the historical 1.23x minimum would require a 54% rise in bitcoin’s price, a 35% decline in network computing power, or a combination of both (21shares, Glassnode, Cambridge Bitcoin Electricity Consumption Index, as of September 30, 2026).
Four developments that could ease pressure on bitcoin in Q3
US Treasury buybacks eased pressure on long-term yields
The US Treasury doubled its long-bond buybacks in September as long-term yields neared multi-decade highs, describing the move as liquidity support. That does not fix rising debt costs, which some market participants argue reinforces the long-term appeal of hard assets such as gold and bitcoin.
The Federal Reserve has room to pause
The Federal Reserve raised rates by 25 basis points in September and projects one more rate hike and then no cuts through 2027. Gasoline drove September's inflation jump, and higher yields already tighten financial conditions. 21shares analysis argues that a pause, with yields and the dollar peaking, would signal a gentler path than markets are pricing in.
Pressure on holders and miners is easing
Net realized losses of $61.6 billion in the first half gave way to $9.3 billion in net gains in Q3, and miner income exceeded full production costs by 6% in September (Glassnode, as of September 30, 2026). The recovery is partial: Q2 and Q3 combined still show a $12.3 billion net realized loss.
Wealth advisors are gaining access to crypto ETPs
Bank of America advisors have been able to recommend crypto exchange-traded products (ETPs) since January, and 32% of surveyed advisors invested for clients in 2025, up from 22%. A 0.1% allocation from the $10.8 trillion US 401(k) retirement market would equal $10.8 billion.
Four risks could reverse the Q3 pattern
Energy price shocks keep yields high
Oil is climbing on Gulf tensions while the Federal Reserve weighs more tightening and heavy borrowing keeps yields high. Although this supports bitcoin as a long-run inflation hedge, higher cash yields give investors a lower-volatility alternative.
Artificial intelligence competes for capital
Earlier in 2026, money rotated from bitcoin ETFs into artificial intelligence (AI) stocks, and miners moved capacity into AI and high-performance computing (HPC). An AI correction would not bring relief, because falling equities could drag bitcoin lower in the short run.
The CLARITY Act stalled, so reform rests on agency action
US regulators are driving reform through agency action while the CLARITY Act has stalled in Congress. That leaves progress exposed to legal challenges and political shifts, and banks may hold back until the rules are settled.
Quantum computing raises a long-run security question
StarkWare has cut the cost of experimental quantum-safe transactions to under $67 from about $320 and more than $300. Coins with lost keys remain a difficult problem, and the issue could erode confidence before any attack occurs.
The Q4 watchlist: demand, supply, and stress signals
ETF demand recovered sharply in Q3, after $4.8 billion of outflows in Q2 when bitcoin fell 14% to ~$58,000. ETFs became the largest identifiable buyer, absorbing some of the bitcoin long-term holders sold into the strength of the rally. That selling was reflected in net realized profit, which swung from a $21.6 billion loss in Q2 to a $9.3 billion gain in Q3, while long-term holders’ share of supply fell from 74.4% to 73%. Trading also shifted further toward futures, whose share of crypto-exchange volume rose from 86.7% to 88.2%.
The next question is whether ETF demand can continue absorbing new supply as selling by long-term holders slows. The five indicators below help track that balance, alongside the growing role of futures. Two further measures, miner profitability and US bond yields, help assess whether the broader conditions remain supportive. Continued ETF inflows above newly mined supply, miners covering their full operational costs, and stabilizing yields would provide a stronger foundation for the recovery
Sources: ETF dollar flows from Bloomberg via 21shares monthly flows report. Other figures are from Glassnode. Data as of September 30, 2026.
FAQ
How much bitcoin did US spot ETFs buy in Q3 2026?
US spot bitcoin ETFs added 79,800 bitcoin (6.2 billion)in Q3 2026, against 41,300 bitcoin ($2.9 billion) newly mined in the same period. The funds had recorded $4.8 billion of outflows in Q2 2026 (Bloomberg, Glassnode, 21shares, as of September 30, 2026).
Who sold bitcoin in Q3 2026?
Long-term holders, meaning wallets whose coins had not moved in more than 155 days, reduced their balances by 248,000 bitcoin ($17.4 billion). Investors overall realized $9.3 billion in net profits, the first quarterly net gain of 2026 (Glassnode, as of September 30, 2026).
Is bitcoin mining profitable in 2026?
Mining was marginally profitable in September 2026, when their income exceeded estimated total mining costs by 6% after ten months below breakeven. Transaction fees covered only 0.67% of miner income in Q3 2026 (Glassnode, Cambridge Bitcoin Electricity Consumption Index, as of September 30, 2026).
